Figure - available from: The Scientific World Journal
This content is subject to copyright.
Trajectory interpolation. (a) Accumulated trajectory on the last frame of gesture “5”; (b) tracked trajectory; (c) interpolated and smoothed trajectory. Columns from left to right are trajectory plot with gesture’s (X,Y); plot X of gesture trajectory; plot Y of gesture trajectory.
Source publication
We propose an adaptive and robust superpixel based hand gesture tracking system, in which hand gestures drawn in free air are recognized from their motion trajectories. First we employed the motion detection of superpixels and unsupervised image segmentation to detect the moving target hand using the first few frames of the input video sequence. Th...
Citations
... e optical camera captures the image sequence of the human motion through the camera, analyzes the human motion characteristics and the motion trajectory in the image sequence, and senses the state of the human body. e method based on optical cameras can be used in many scenes, such as gesture recognition [15], gait recognition [16], and target tracking [17]. However, this method still has shortcomings, and it cannot work in lowlight conditions and where privacy is involved. ...
WiFi indoor personnel behavior recognition has become the core technology of wireless network perception. However, the existing human behavior recognition methods have great challenges in terms of detection accuracy, intrusion, and complexity of operations. In this paper, we firstly analyze and summarize the existing human motion recognition schemes, and due to the existence of the problems in them, we propose a noninvasive, highly robust complex human motion recognition scheme based on Channel State Information (CSI), that is, CSI-HC, and the traditional Chinese martial art XingYiQuan is verified as a complex motion background. CSI-HC is divided into two phases: offline and online. In the offline phase, the human motion data are collected on the commercial Atheros NIC and a powerful denoising method is constructed by using the Butterworth low-pass filter and wavelet function to filter the outliers in the motion data. Then, through Restricted Boltzmann Machine (RBM) training and classification, we establish offline fingerprint information. In the online phase, SoftMax regression is used to correct the RBM classification to process the motion data collected in real time and the processed real-time data are matched with the offline fingerprint information. On this basis, the recognition of a complex human motion is realized. Finally, through repeated experiments in three classical indoor scenes, the parameter setting and user diversity affecting the accuracy of motion recognition are analyzed and the robustness of CSI-HC is detected. In addition, the performance of the proposed method is compared with that of the existing motion recognition methods. The experimental results show that the average motion recognition rate of CSI-HC in three classic indoor scenes reaches 85.4%, in terms of motion complexity and indoor recognition accuracy. Compared with other algorithms, it has higher stability and robustness.
... And the ones working with temporal gestures are computationally expensive for a mobile device. Zhu and Pan [55] employ superpixel based image segmentation(SLIC) and use temporal shifts in the obtained superpixels to model hand structure Superpixels can be extremely heavy for a smartphone to process and add a significant amount of latency to the detection. Hegde et al. [19] Chapter 3 ...
... Methods like Grabcut [43,19] and Superpixel based segmentation [55] were effective, but very slow in case of realtime computing on a mobile device. The color based approach presents a simpler algorithm, which works with RGB channel stream for smartphone monocular camera without built-in depth sensors. ...
The future of UI will be dominated primarily by gestures. With the rise in the popularity of platforms like Augmented Reality(AR) and Virtual Reality(VR), accompanied by the rapid growth in computational ability of smartphones and wearables, real-time gesture recognition is one of the prime research topics in Human-Computer Interaction(HCI). While there are quite a few existing methods for gesture recognition, very few of them address the challenges involving real-time performance on smart-phones with comparatively low computational ability and egocentric vision. In this work, I explore the numerous existing techniques, widely accepted across the research community, and study the challenges involved in porting them to First Person View (FPV). The techniques cover a wide set of areas, ranging from band-pass filtering to state of the art deep learning algorithms and includes a novel approach, proposed as a computationally inexpensive algorithm to detect hand swipes from a smartphone. The main motive is to study the feasible methods available for real-time gesture recognition under FPV constraints.
... By analyzing the patches with landscape, the BHMC enhanced smooth-ness and steepness properties of local mode. Recently, many works [9][10][11][12][13] have been also proposed to address the problem of object segment and tracking. ...
Video object tracking is one of the most important topics in computer vision. There are many applications in visual tracking technologies, for instance, surveillance, augmented reality, gesture recognition and interactive gaming. In this paper, we proposed a novel method for tracking highly non- rigid objects by over-segmentation and statistical learning. Rather than using conventional bounding box, the tracker is based on segments and considers the target object to be a combination of segments. Objectness method is employed and integrated into the tracker to generate candidates for similarity measure. Moreover segment-based motion weights are introduced to give higher weights on candidates that in motion consistency. A confidence-collecting scheme is proposed for similarities of candidates. To validate our method, experiments are conducted on several image sequences with different non-rigid challenges. The experimental results are satisfactory for highly non-rigid object tracking.
A novel scheme for efficient shape classification using region descriptors and extreme learning machine with kernels is proposed. The skeleton and boundary of the input shape image are first extracted. Then the boundary is simplified to remove noise and minor variations. Finally, region descriptors for the local skeleton, and the simplified shape signature are constructed to form a hybrid feature vector. Training and classification are then performed using kernel extreme learning machine (k-ELM) for efficient shape classification. Experimental results show that the proposed scheme is very fast and can archive higher classification accuracy on the challenging MPEG-7 dataset, outperforming existing state-of-the-art methods.
This thesis presents Variable Reality, a wearable augmented reality-based system focused on creating a unique on-the-go reading experience that combines the readily accessible nature of digital books with the favorable physical spatiality of a paper book. The two types of Variable Reality books are the Augmented Book and the Virtual Book. They differ in the way they are displayed as the former augments virtual pages onto an actual book whereas the latter virtually augments a 3-D book on the palm of the user's hand. Designed to take the physical form of a book, Variable Reality books make use of human cognitive ability in storing and retrieving information in a spatial manner. Easy-to-use hand gestures naturally associated with reading activity are integrated with the system to help bring an intuitive user experience in reading.
